def copy(self):
"""need to reimplement this because the __init__ signature of
IntDict is not identical to that of defaultdict"""
result = IntDict(self.parent)
defaultdict.update(result, self)
# see http://www.logilab.org/ticket/23986
return result
def __init__(self, item_coder_data=None, properties=None,
boundary_types=None, boundary_format=BoundaryFormat.mass):
'''
Initialize a dataset.
'''
defaultdict.__init__(self, dict)
self.properties = dict()
self.boundary_format = boundary_format
# Masses
if item_coder_data is not None and item_coder_data is not dict:
defaultdict.update(self, item_coder_data)
# Properties
if properties is not None:
self.properties.update(properties)
# Boundary types
if boundary_types is not None:
self.boundary_types = set(boundary_types)
else:
self.boundary_types = set([1])
# Coders
self.coders = set()
# Populate coders
for coder_masses in defaultdict.values(self):
for coder in coder_masses.keys():
self.coders.add(coder)
def runTrails(self,
alphas=[0.001, 0.005, 0.00592],
betas=[0.0003, 0.0005, 0.00798],
lambdas=[0, 0.2, 0.4, 0.6],
degree=[3],
path="Result/"):
print("actor_critic(lambda)\n")
log = "alpha beta lambda2 degree mean_return max_return last_return\n"
dict = {}
for alpha in alphas:
for beta in betas:
for lambda2 in lambdas:
trails = []
for i in range(self.trail_n):
if self.state_type == "continue":
result = self.run(alpha=alpha,
beta=beta,
lambda2=lambda2)
else:
result = self.run_Tabular(alpha=alpha, beta=beta)
trails.append(result)
path_temp = path + self.env_name + "/" + self.agent_name + "/"
name = path_temp + str(alpha) + "_" + str(
beta) + "_" + str(lambda2)+"_"+\
str(self.degree)
name = name.replace(".", "")
pickle.dump(trails, open(name, "wb"))
plot_trails(trails, name)
trails = np.array(trails)
trails = trails[:, 90:]
dict.update({
str(alpha) + "_" + str(beta) + "_" + str(lambda2) + "_" + str(self.degree):
np.mean(trails)
})
print("done one trails", np.mean(trails))
sorted_J = sorted(dict.items(),
key=operator.itemgetter(1),
reverse=True)
print(str(sorted_J[0]))
f = open(path_temp + 'log.txt', 'a')
f.write(str(sorted_J))
f.write("\n")
f.close()
def runTrails(self,
alphas=[0.001, 0.002, 0.003, 0.008],
epsilons=[0.008, 0.01, 0.05],
lambda2s=[0, 0.3, 0.5, 0.8],
path="Result/"):
log = "alpha epsilon lambuda degree mean_return max_return last_return\n"
dict = {}
for alpha in alphas:
for epsilon in epsilons:
for lambda2 in lambda2s:
trails = []
for i in range(self.trail_n):
if self.state_type == "continue":
result = self.run(alpha=alpha,
epsilon=epsilon,
lambda2=lambda2)
else:
result = self.run_tabular(alpha=alpha,
epsilon=epsilon,
lambda2=lambda2)
print("done one", i)
trails.append(result)
path_t = path + self.env_name + "/" + self.agent_name + "/"
name = path_t + str(alpha) + "_" + str(
epsilon) + "_" + str(lambda2) + "_" + str(self.degree)
name = name.replace(".", "")
pickle.dump(trails, open(name, "wb"))
plot_trails(trails, name)
trails = np.array(trails)
trails = trails[:, 90:]
dict.update({
str(alpha) + "_" + str(epsilon) + "_" + str(lambda2) + "_" + str(self.degree):
np.mean(trails)
})
print("done one trails", np.mean(trails))
sorted_J = sorted(dict.items(),
key=operator.itemgetter(1),
reverse=True)
print(sorted_J[0])
f = open(path_t + 'log.txt', 'a')
f.write(str(sorted_J))
f.write("\n")
f.close()
